Silencing of TESTIN by dense biallelic promoter methylation is the most common molecular event in childhood acute lymphoblastic leukaemia

To identify genes showing frequent high-level methylation, DNA from five ALL marrow samples and one peripheral blood sample were analysed. The density of methylation within 24 gene promoter regions was determined by using MS-MLPA, in multiple independent experiments. Percent methylation at the HhaI site was calculated by comparing normalised, HhaI-digested peak areas to normalised, mock-digested peak areas.

To validate the MS-MLPA results, detailed analysis of TES promoter (Additional file 1, Figure S1) methylation was obtained by performing bisulfite sequencing. ALL DNA samples and normal PBL DNA were bisulfite treated, amplified using bisulfite-specific primers, cloned and sequenced. Each of the ALL samples was hypermethylated compared to peripheral blood (Figure 1A). ALL5 (T lineage; Age: 7 years 7 months), for which methylation was not detected by MS-MLPA, showed the lowest level of methylation, but was methylated at CpG sites other than the HhaI site interrogated by MS-MLPA.

Bisulfite sequencing was then performed on a larger group of samples including normal peripheral blood, normal bone marrow, additional ALL bone marrow samples and remission bone marrow samples. Normal peripheral blood (n = 5; Figure 1B) and bone marrow (n = 6; data not shown) samples showed no methylation of the TES promoter. Leukaemia DNA, but not matched remission samples showed methylation of the TES promoter (Figures 1C and 1D). In summary, twenty out of twenty ALL samples were hypermethylated compared to normal blood and bone marrow at the TES promoter. Five of the ALL samples showed a small number of completely unmethylated alleles. These alleles could reflect the presence of a population of non-malignant cells in the marrow aspirate sample. For ALL samples, the percentage of methylation for all sites in the promoter ranged from 11% to 86% with a median of 65% and an interquartile range of 50-74%. In particular, B ALL samples had median methylation of 71% (interquartile range of 64-77%), whilst the T ALL samples (ALL5, ALL13 (6 years 1 month) and ALL15 (5 years 3 months)) had median methylation of 32.9% (B ALL vs. T ALL: p < 0.01 by T test). In contrast the percentage methylation with peripheral blood ranged from 0 to 1.5%. Since the amplification and cloning of bisulfite-treated DNA is subject to biases, the results are only semi-quantitative.

To confirm that the observed hypermethylation did not reflect preferential cloning biases within the bisulfite conversion-derived PCR products, we developed a combined bisulfite restriction assay (CoBRA). After bisulfite conversion, reverse strand amplification of fully methylated DNA generates four Taq^αI sites (TCGA) at CpG sites 3, 10, 16 and 30, whereas none are present in unmethylated DNA after amplification. CoBRA was performed on six ALL samples, one matched remission and one normal PBL sample. In all cases, the validity of the bisulfite sequencing results was confirmed by the CoBRA analysis (Figure 2). For example, ALL8 (B lineage; 5 years 9 months) and ALL12 (B lineage; 5 years), which show dense methylation by bisulfite sequencing (Figure 1C), showed a CoBRA banding pattern indicative of dense methylation. Similarly, the partial methylation shown by ALL19 (B lineage; 6 years 3 months) (Figure 1D) is paralleled by a CoBRA banding pattern indicative of partial methylation at CpG10 and CpG16. The unmethylated remission and normal PBL were undigested by Taq^αI, also in agreement with bisulfite sequencing results.

The presence of SNPs (rs1319886, rs28411392 and rs11549785) in the TES promoter permitted analysis of allele-specific methylation. Bisulfite sequencing of the reverse strand confirmed biallelic methylation in five informative ALL samples (Table 1). Additionally the presence of another promoter SNP (rs11549786), outside of the bisulphite PCR amplicon, and of a telomeric microsatellite (D7s655) permitted copy number analysis. In 15 out of 20 cases LOH was excluded, but in the remaining 5 non-informative cases LOH could not be excluded.

TES mutations have been reported in T ALL (CCRF-CEM), breast cancer and ovarian cancer cell lines [18, 19] and we hypothesised that the less methylated ALL samples may contain coding mutations. Exon sequences from ALL5 and ALL19 (the two samples with the lowest density of methylation) were amplified and sequenced. No mutations were found (data not shown), however we were unable to exclude the possibility of intragenic deletions not readily detectable by amplification and sequencing.

TES promoter methylation has been shown to down-regulate expression in breast cancer cell lines, leukaemia cell lines [18] and glioblastomas [20]. To determine the role of methylation on TES expression in ALL we used microarrays to compare expression in a separate cohort of paediatric precursor B ALL specimens with two control precursor cell populations, normal bone marrow CD34+ progenitor cells (n = 5) and umbilical cord blood CD19+IgM- (pre-B) cells (n = 3). Leukaemia samples were classified into subtypes using expression profiling based on Yeoh et al. [21].

As previously reported, in a comparison of gene expression between normal bone marrow CD34+ cells and precursor B ALL, TES was among the most down-regulated genes as assessed by Affymetrix HG-U133A microarrays [22]. Under a linear regression model a marked reduction in relative expression of TES (probe-set 202719_s_at) was observed in the common subtypes of precursor B ALL, including hyperdiploid, TEL-AML1, BCR-ABL, E2A-PBX1 (for each subgroup p < 0.001) and in T lineage ALL (p < 0.05) but not in MLL rearranged ALL (p = 0.20), when compared to CD34+ BM (Figure 3). Similar p-values were recorded when CD19+ CB was used as the comparison or if the second TES probe-set (202120_at) was used, though p-values were not always as strong. In total marked reduction in TES expression was observed in 68 of the 74 cases of B ALL (Figure 3).

The relationship between TES expression and methylation was examined directly in available leukaemia cell lines. DNA methylation status was determined using CoBRA as before (Figure 4A). In brief, MOLT4 (T ALL) bisulfite-specific PCR product was completely digested by Taq^αI, indicating that MOLT4 cells are methylated at the four CpG sites interrogated by Taq^αI. Raji (Burkitt lymphoma) cells appear to be hemi-methylated, i.e., methylated on one allele and unmethylated on the other allele. RT-PCR was performed on total cell line RNA using exon-specific primer pairs. PCR products from amplifications with primers spanning exon 3 and exon 6 are shown (Figure 4B). TES expression was present in Raji and normal PBL, but was undetectable in MOLT 4 cells by RT-PCR. Also, TES expression levels were quantified using real-time quantitative RT-PCR (qRT-PCR) with expression levels calculated relative to normal PBL levels after normalisation to β2-microglobulin and reproduced on Figure 4B. TES promoter methylation resulted in a reduction of TES expression in the leukaemia cell lines tested in agreement with previously published reports [18, 20].

To confirm the inverse relationship between methylation and expression, CoBRA and qRT-PCR was performed on TES expressing, myeloid-lineage cell lines: HL60, K562 and U937. Bisulfite-specific PCR products from HL60, K562 and U937 cells were not digested by Taq^αI, thus demonstrating the inverse relationship between methylation and expression.

As previously reported [18] multiple splice variants of TES were detectable by RT-PCR; however an unexpected splice variant was amplified from both cell lines and normal blood (Figure 4B). This novel splice variant was sequenced and compared with published variants. The novel variant cDNA uses a cryptic splice site present in exon 5 to generate a 102 bp shorter cDNA product, which if translated would result in a truncated protein of 237 amino acids (see Additional file 2, Figure S2).

Next, fourteen ALL xenografts [23] were analysed for TES promoter methylation and expression analysis. Methylation status was measured by CoBRA assay and was in complete, reciprocal agreement with expression levels measured by real time RT-PCR (see Figure 4C). Without knowledge of their phenotypes, we predicted that almost all methylated and non-expressing xenografts would be B-lineage derived, whereas unmethylated and TES expressing xenografts would be T-lineage ALL. xALL 2, 3, 4, 7, 8, 10, 11, 19 and 30 were predicted to be B-lineage: this was correct except for xALL 30 which was T ALL. xALL 16, 17, 27, 29 and 31 were predicted to be T ALL: xALL 16, 27, 29 and 31 were confirmed to be T-lineage ALL, whereas xALL 17 was B-lineage [23]. Analysis of the karyotype of xALL 17 (46, XX, -20, +21 (18), XX, -20, +21, +mar (4)) suggests that xALL 17 be classified as "Other" according to Yeoh et al. [21] (see Figure 3). Of the TES- expressing xenografts, xALL31 appeared to show hemi-methylation of the TES promoter with 50% of the bisulfite-specific PCR product not being digested by Taq^αI. This hemi-methylation was similar to that seen with the Raji cell line (see Figure 4A). In conclusion, we were able to predict the lineage of ALL xenografts on the basis of TES expression and methylation.

We have demonstrated TES hypermethylation in 17 of 17 B ALL samples using bisulfite sequencing and down-regulation of TES expression using expression profiling in 68 of 74 independent B ALL samples. In addition, we have demonstrated both promoter methylation and silencing of TES expression in 8 of 9 B ALL xenografts. In total, transcriptional silencing and/or promoter methylation of TES occurred in 93 of 100 B ALL and 20 of 28 T ALL cases.

Paediatric acute lymphoblastic leukaemia (n = 20) and matched remission (n = 5) bone marrow aspirate samples, normal marrow and normal peripheral blood leukocyte (PBL) samples were collected in accord with ethical approval obtained from the Otago Ethics Committee. All samples contained at least 80% blasts, and usually at least 90%. The remission samples that were used were obtained at least 35 days after diagnosis. The median age of the ALL patients was 4 years 10 months (range 2 years to 11 years 3 months), and comprised 17 B lineage and 3 T lineage (ALL5, ALL13 and ALL15) ALL. Samples were obtained prior to routine molecular testing; two B lineage cases showed a hyperdiploid karyotype (ALL11 and ALL14); one B lineage case (ALL2) had t(9;22) and no cases showed structural 11q23 abnormalities (MLL gene location).

Cell lines were newly obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and maintained according to ATCC instructions.

A separate group of ALL samples were maintained as xenografts in SCID/NOD mice as previously described [23].

Twenty-four candidate genes were chosen from published literature. At the time of selection (2005) these candidates included the majority of genes known to be methylated in haematopoietic cancers.

MS-MLPA probes were designed to interrogate HhaI sites within the reported region of methylation and purchased from Sigma-Proligo (St Louis, MO, USA)(see Additional file 4, Table S4). The 24 HhaI site containing probes and three control probes were prepared as three separate probe-mixes and used with the MLPA EK1 kit (MRC Holland, Amsterdam, The Netherlands). MS-MLPA assays were performed as per manufacturer's instructions. After probe ligation each sample was divided in two, with one aliquot being incubated with HhaI. Ligated probes were then amplified using MLPA-universal primers and products were detected with an ABI3100 capillary sequencer, and analysed using GENESCAN (Applied Biosystems Ltd, Foster City, CA, USA) and EXCEL (Microsoft, Redmond, WA, USA) software. Three control probes, not digested by HhaI, were used for normalisation of probe peak areas.

Each probe peak area was normalised using the sum of the control probe peak areas. Percent methylation was calculated from the normalised HhaI-digested peak area and the normalised mock-digested peak area (as below).

DNA from ALL and matched remission samples, normal PBL and bone marrow samples were bisulfite-treated using either the EZ DNA Methylation Gold system (Zymo Research, Orange, CA, USA) or our standard bisulfite-treatment protocol [46].

Bisulfite-specific primers were designed using the MethPrimer website [47] (Forward: ^5' TTAGGGTTATTGAGTTTGTTTAGTAGG ^3'; Reverse: ^5' CTTTATTTTCCAAATCCATATTAAC ^3'), and were used to amplify 371 base pairs or 48 CpG dinucleotides of the TES promoter. Amplified products were cloned using TOPO TA Cloning system (Invitrogen, Carlsbad, CA, USA) and sequenced using ABI BigDye 3.1 Terminator System and supplied TOPO M13 reverse primer. Resulting sequences were viewed and, if necessary, edited in 4 Peaks http://​mekentosj.​com/​4peaks/​, and aligned using the Se-Al v2.0a11 Carbon program http://​evolve.​zoo.​ox.​ac.​uk/​. Percent methylation was calculated as the percentage of methylated CpGs out of total CpGs sequenced.

Amplification of genomic DNA was performed using TES promoter-specific primers (Forward: ^5' ACCAGGTCAGGGTCACTGAGCTTGC ^3'; Reverse: ^5' ACCCGCGCAGGTGAAGCAGC ^3') to investigate four SNPs (rs11549786, rs11549785, rs28411392 and rs1319886). PCR products were purified using DNA Clean-Up kit (Zymo Research) and sequencing was performed as above, using genotyping PCR primers.

CoBRA was designed to interrogate four Taq^αI sites generated from methylated DNA after bisulfite conversion. DNA was bisulfite-treated and amplified using primers designed to the reverse strand (Forward: ^5'ATTTTGTTTTTTAGGTTTATGTTAA ^3'; Reverse: ^5' CCAAATCAAAATCACTAAACTTACC ^3'). After PCR amplification and DNA Clean-Up (Zymo Research), purified products were digested with Taq^αI and electrophoresed through 2% Seakem LE agarose (Lonza, Basel, Switzerland). Amplified products were cloned and sequenced as above, to generate SNP genotyping data.

Total RNA was extracted from 101 bone marrow specimens from paediatric ALL patients, from five isolates of CD34+ cells from normal marrow, and from three isolates of CD19+IgM- cells from umbilical cord blood as previously described [22]. cRNA was prepared and hybridised to HG-U133A microarray (Affymetrix, Santa Clara, CA, USA) as described [22]. ALL samples were classified into leukaemia sub-groups, as described by Yeoh et al. [21]. TES expression was calculated from both TES-specific probes and median expression level for each ALL sub-group was calculated.

First strand cDNA was generated from total RNA using Superscript III Reverse Transcriptase (Invitrogen) and manufacturer's instructions, except a mix of oligo-dT and random primers were used to prime the synthesis. Multiple exon-specific primers were designed to amplify cDNA, with genomic DNA not being amplified or producing larger sized products. RT-PCR products were cloned using the TOPO TA Cloning System before being sequenced. TES expression levels were measured using ABI Assay-On-Demand kits (TES: Hs_00932509_g1; β2-microglobulin: Hs00984230_m1) and ABI PRISM 7900 HT Sequence Detection System (Applied Biosystems Ltd) according to manufacturer's instructions. TES expression levels were normalised to β2-microglobulin before calculation relative to normal PBL levels.